Research On Image Mosaic

Image mosaic is a technology that carries on the spatial matching to a series of image which are overlapped with each other, and finally builds a seamless and high quality image which has high resolution and big eyeshot compared with a single image. Image mosaic is an important research field of image processing, and has widely applications in the fields of photogrammetry, computer vision, remote sensing image processing, medical image analysis, computer graphic and so on.The thesis first introduces the present research status of image mosaics technology, and the characteristic and application domain of mosaics, thus has demonstrated the broad application prospect of image mosaics technology. Then the thesis discusses the basic steps of image mosaic, image acquisition methods and models of image mosaic. Moreover, it researches with emphasis in the two core technology of image mosaic, which are the image registration technology and the image fusion technology. The common image registration methods and image fusion methods are introduced, and their performance as well as the existing problems is analyzed. The thesis also discusses the panoramic image technology, which is a popular application of image mosaic technology. The panoramic image technology has found widely used in the construction of virtual environments and virtual travel. Then the three types of panoramic image are analyzed. The main task of this paper is the research of feature-based method, on the basis of this, a novel stitching algorithm based on direction of image feature for images with formation of translation and rotation is proposed. The intensity information and position information of interest point is utilized in this algorithm. Interest points are firstly extracted by the Harris angle detector in this method, and initial matches are extracted by the method of dominant orientation and cross correlation, then the false matches are rejected according to the matching degrees of feature points. Transform parameters are determined by the matches, and weighted average method is used to blend images. Experimental results show that the proposed algorithm is accurate in matching and having good capabilities.